| Summary: | 碩士 === 國立交通大學 === 電信工程研究所 === 100 === Two main issues involved with the performance of multi-channel multi-radio (MCMR) wireless mesh networks (WMNs) are channel assignment (CA) and multi-channel routing (MCR). Joint CA and MCR problem has been proven to be NP-hard. In this thesis, we propose to apply genetic algorithms for CA problem and tackle MCR using linear programming techniques. Since CA and MCR tend to interact with each other, to reflect such interplay property, we evaluate the fitness value of a chromosome (certain CA configuration) in our genetic algorithms by computing the linear objective function. Therefore, we successfully decouple the two problems and obtain an optimized CA configuration with corresponding MCR schedule in polynomial time at the WMN planning stage. In our implementation, we design a router software the incorporates our WMN resource management algorithm to maximize the system throughput. When evaluating the software on our operational testbed, we observe three factors that prevent the resource management algorithm from performing well as expected. The first factor is the inaccuracy of wireless link capacity estimation, the second one is the imperfection of traffic scheduling, and the third one is the difficulty of satisfying pre-defined user bandwidth demand. Consequently, the computed routing path is not optimized as the algorithm intends to. While the three factors are generally assumed to be solvable in theoretical optimization models, from our prototyping experiences, we conclude that these assumptions do not always hold in practical networks.
|
|---|
